Credit Risk Prediction

Developed an end-to-end machine learning system to predict customer credit risk using real-world financial and demographic data. The project involved data preprocessing, feature engineering, model training, and evaluation to classify customers as low or high credit risk.

The dataset consisted of over 400,000 real customer records, including personal information, income, employment history, and credit payment behavior. Extensive data cleaning was performed to handle missing values, outliers, and inconsistent records. Additional features such as customer age and credit history metrics were engineered to improve model performance.

A classification model was trained and evaluated using train-test split and cross-validation to ensure generalization and prevent overfitting. The model achieved:

Train Accuracy: 97.05%

Test Accuracy: 97.07%

Cross-Validation Accuracy: 95.84%

The close performance between training and testing results indicates strong generalization and minimal overfitting.

This system can be used in real-world financial applications such as:

Credit approval systems

Loan risk assessment

Banking risk management

Automated credit scoring systems

Technologies Used

Python

pandas for data processing

NumPy for numerical operations

scikit-learn for machine learning

Matplotlib for visualization